Technical Field
[0001] The present invention relates to communications networks, in general, and in particular
to multi-domain communications networks.
Background
[0002] Next Generation Networks are large, complex and heterogeneous structures and the
relative control plane should automatically and dynamically provide end-to-end connections
spanning domains, technology regions and layers.
[0003] The main standardization bodies (IETF, OIF and ITU) are addressing several aspects
of routing in a multi-domain/multi-technologies scenario, where a single domain can
also be multi-layer (i.e. a Wavelength-Switched Optical Network, WSON, domain with
packet interfaces at its borders). The common approach followed for multi-layer, multi-technology
and multi-domain networks is mainly based on the division of the network resources
into "technology regions" (e.g. optical and packet regions), having hierarchical server-client
relations. On such architecture, the Generalised Multiprotocol Label Switching, GMPLS,
protocol suite, with the relative extensions for each technology and layer, should
allow a homogeneous control of portions of the network, which are heterogeneous both
technologically and administratively.
[0004] The large amount of network elements and their management complexity imply a set
of scalability issues both for routing and for signaling protocols. As a matter of
fact, the amount of information to be distributed and the frequency of updates increase
the overhead of the entire network signaling and thus the bandwidth request and the
computational power consumed. On the other hand, poor information (e.g. Traffic Engineering,
TE, information not transmitted from one hierarchical layer to another one or a too
generic resource virtualization) can lead to an inefficient resources utilization
and to a high blocking probability (i.e. the probability that a connection seen as
feasible at a higher hierarchical level can not be provided, because of a lack of
resources in the lower level).
[0005] Moreover, different domains can be administrated by different carriers as well, resulting
in a set of confidentiality issues due to strong commercial reasons that refrains
a domain administrator to disclose all the details of its intra domain topology.
[0006] However a common view of these complex topics has not been proposed so far and a
set of issues are still to be fixed.
[0007] As a first step for providing a framework for hierarchical architectures with resources'
summarization a patent application
WO2011/103913A1 has been filed, defining a general hierarchical architecture with service-oriented
parameters. This patent application addresses the issue of defining a general architecture
and a common set of parameters, but it does not specify any method to perform resource
summarization.
[0008] In order to fix the scalability issue in multi-region and multi-carrier scenarios,
the design of hierarchical architectures, based on Path Computation Elements (PCEs),
seems to be a promising approach, currently under standardization.
[0009] However, such approach is not completely defined and technology-specific extensions
are still needed. Moreover, End-To-End, E2E, path computation procedures should be
further improved and extended in order to work efficiently also on summarized topologies.
[0010] Concerning the topology summarization concept itself, a set of guidelines are expressed
in the standards, but they only indicate that a domain can be summarized as a single
node, as a full mesh of links interconnecting some border nodes, or some generic intermediate
solution. Moreover, the domain summarization is actually mainly considered as an enabler
for preliminary considerations such as the domains' chain definition, not exploiting
its real potential.
[0011] Another important lack of the existing solutions is that, for both PCE hierarchical
architectures and domains summarization, standards only provide guidelines and general
approaches. As a matter of fact, there is no common view with any specific solutions
or defined framework addressing all the issues of Multi-Layer, Multi-Region, Multi-Vendor
and Multi-Domain Networks.
[0012] As mentioned in the previous section, the patent application
WO2011/103913A1 has been filed as a first step for providing a framework for hierarchical architectures
with resources summarization. It describes a general hierarchical architecture to
summarize domain topologies in a multi-domain and multi-technology context to be used
for E2E solutions. Main concepts of
WO2011/103913A1 are:
- Architectural definition of specialized PCEs, summarizing the intra-domain topologies,
and inter-domain external PCE, working on a summarized view of the topologies.
- Definition of service-based parameters to be used for the summarization of each domain,
acting as "umbrella" parameters for different technologies.
[0013] WO2011/103913A1 addresses the issue of defining a general architecture and a common set of parameters;
however it does not specify any method to perform an efficient resource summarization.
Summary
[0015] The present invention provides a specific solution for communications network topology
summarization with a good trade-off between high scalability, good stability and resources
utilization efficiency.
[0016] These features are very important added values for E2E connectivity provisioning
in Multi-Layer, Multi-Region, Multi-Vendor and Multi-Domain Networks. As a matter
of fact, in such contexts, is not trivial to find a good trade-off between the level
of abstraction (improving scalability), the stability (reducing the amount of signaling)
and the adequate level of detail for optimizing resources' utilization and reducing
the blocking probability (due to failed attempts to setup a connection based on non
consistent topology information). The invention seeks to preferably mitigate, alleviate
or eliminate one or more of the disadvantages mentioned above singly or in any combination.
[0017] Accordingly, the invention seeks to preferably mitigate, alleviate or eliminate one
or more of the disadvantages mentioned above singly or in any combination.
[0018] According to a first aspect of the present invention there is provided a method of
building a summarised topology of a communications network. The method comprises the
steps of receiving Class of Service, CoS, settings defining a plurality of CoS baskets,
each CoS basket being defined by a specific combination of a range of delay and a
range of bandwidth. The method also comprises receiving traffic information and traffic
demands for traffic between pairs of border nodes. The method also comprises building
an Expected Traffic Matrix, EM, based on the received Class of Service settings, traffic
information and traffic demands, such that each traffic demand, which is specified
by a source, a destination and bandwidth, is associated to a basket compatible with
the features of said traffic demand and routing traffic demands represented in the
Expected Traffic Matrix; creating a path and assigning the path to one of CoS baskets
for traffic demands in the EM matrix, where the CoS baskets represent different Classes
of Service. If a traffic demand can be served then a path in the basket with the smallest
bandwidth greater than the bandwidth of the traffic demand is chosen for assigning
to the traffic demand and the residual bandwidth is made available future reservations.
Furthermore the method comprises associating each CoS basket with a Summary Link connecting
a pair of border nodes, wherein the Summary link represents a set of connections and
building a summarised topology of the network, wherein pairs of border nodes are linked
by the Summary Links corresponding to the CoS baskets.
[0019] According to a second aspect of the present invention there is provided a device
for building a summarised topology of a communications network. The device comprises
an interface adapted to receive Class of Service, CoS, settings defining a plurality
of CoS baskets, each CoS basket being defined by a specific combination of a range
of delay and a range of bandwidth. The interface is also adapted to receive traffic
information and traffic demands for traffic between pairs of border nodes. The device
further comprises a processor adapted to build an Expected Traffic Matrix, EM, based
on the received Class of Service settings, traffic information and traffic demands,
such that each traffic demand, which is specified by a source, a destination and bandwidth,
is associated to a basket compatible with the features of said traffic demand and
route traffic demands represented in the Expected Traffic Matrix, EM. The processor
is further adapted to create a path and assign the path to one of CoS baskets for
traffic demands in the EM matrix, where the CoS baskets represent different Classes
of Service. If a traffic demand can be served then the processor is further adapted
to choose a path in the basket with the smallest bandwidth greater than the bandwidth
of the traffic demand for assigning to the traffic demand and to make the residual
bandwidth available for future reservations. The processor is also adapted to associate
each CoS basket with a Summary Link connecting a pair of border nodes, wherein the
Summary link represents a set of connections. The device further comprises a topology
module adapted to build a summarised topology of the network, wherein pairs of border
nodes are linked by the Summary Links corresponding to the CoS baskets.
[0020] Further features of the present invention are as claimed in the dependent claims.
Brief description of the drawings
[0021] The present invention will be understood and appreciated more fully from the following
detailed description taken in conjunction with the drawings in which:
FIG. 1 is a diagram illustrating two phases of the method in one embodiment of the
present invention;
FIG. 2 is a diagram illustrating construction phase of the method in one embodiment
of the present invention;
FIG. 3 and Fig. 4 illustrate CoS baskets in one embodiment of the present invention;
FIG. 5 is a diagram illustrating service and update phase of the method in one embodiment
of the present invention;
FIG. 6 is a diagram illustrating a method of building a summarised topology of a communications
network in one embodiment of the present invention;
FIG. 7 is a diagram illustrating a device for building a summarised topology of a
communications network in one embodiment of the present invention;
FIG. 8 is a diagram illustrating a communications network using a device for building
a summarised topology in one embodiment of the present invention.
Detailed description
[0022] The purpose of the current invention is to define an efficient method to provide
a summarized domain topology, allowing an efficient routing in multi-domain contexts
with heterogeneous technologies such as packet and optical.
[0023] Particularly, a detailed method is provided, in order to summarize the domain's connectivity
by a set of baskets, representing different Class of Services (CoS), which are described
with service-oriented and economic-driven parameters.
[0024] Moreover, the intra-domain paths for filling the CoS baskets are computed and clustered
in innovative and efficient way, with the goal of achieving a good trade off between
the following contrasting targets:
- High scalability: this target is achieved by representing a large set of connections
between a given couple of border nodes, BNs, as a small number of Summary Links, with
different characteristics. This way, the topology of a domain is kept small and, at
the same time, its connectivity capabilities are clustered into different Class of
Services (CoS);
- Stability (i.e. robustness to network resources status): this target is achieved by
a set of sub-sequential expedients. First, the intra-domain paths for filling the
baskets are chosen having the minimum (ideally no) mutual influence, so that the resources
of one path don't impact other ones. As a second expedient, the parameters to be advertised
correspond to a large set of intra-domain paths, so that the advertised parameters
would not be updated even if a large number of similar connections are requested.
Finally, when crossing a given threshold for the number of still available paths corresponding
to the advertised parameters, a specific path computation is triggered, in order to
refill (if possible) the baskets instead of modifying the advertised parameters ;
- Confidentiality: this target is achieved by hiding the number, type and Traffic Engineering,
TE, characteristics of the intra-domain paths. As a matter of fact, only the availability
of an undefined set of connections, with some delay and bandwidth characteristics
is advertised, with no details about their composition, TE parameters, etc. ;
- Resources' utilization efficiency: this target is achieved by providing an adequate
level of detail of the service-oriented characteristics (i.e. bandwidth and delay).
That allows the knowledge of some surely met requirements (i.e. the guaranteed bandwidth
and the E2E delay range advertised for each SL).
[0025] As a result, given a domain with its topology and available resources, it is possible
to build a scalable, stable and efficient mesh topology of Border Nodes (BNs), interconnected
by Summary Links (SLs), each one of them representing a whole set of connections.
[0026] Moreover, such summarized topology and corresponding service parameters can be provided
taking into account intra-domain policies and proper physical constraints (e.g. all
connections can be pre-computed in advance, partially computed in advance or completely
computed real time). The internal parameters used for intra-domain routing can be
physical impairments (in case of optical networks), or others. In this way inter-domain
topology is "quasi-static" and based on homogenous parameters, while intra-domain
routing is completely independent from other domains and is based on different parameters
(e.g. available bandwidth, network load, etc.) with respect to inter-domain routing.
[0027] The proposed summarization method can be divided in two phases: a Summarized Topology
Construction phase and a Service & Update phase.
[0028] As shown in Fig. 1 flowchart, given a domain topology 102 and the settings 104 related
to the desired Class of Services (CoS), the Construction phase 106 interacts with
the domain Internal PCE (I-PCE) 108 and the Traffic Engineering Database, TE DB, 110
in order to construct and fill the CoS Baskets 112, which will be advertised as Summary
Links (SL) 118 with service-oriented and cost-based parameters. Given a service request
114, the Service & Update phase 116 selects the adequate path 120 belonging to the
relevant CoS (i.e. Basket) and, if needed, updates the basket database 112 and/or
the parameters advertised for the summarized topology 118. In the following, the two
phases will be explained in details. The advertised Summary Links 118 are virtualisations
of real intra-domain paths. From interdomain point of view the invention in its embodiment
considers links connecting border nodes, but such links represent physical paths that
cross physical nodes into each domain.
Summarized Topology Construction phase
[0029] As shown in the flow chart of Fig. 2, describing this phase of the summarization
method, an Expected Traffic Matrix, EM, is build, based either on an optional traffic
matrix 202 provided as input or stochastic distributions. Each traffic demand in EM
is between border nodes and comes with a bandwidth and a delay request.
[0030] The demands for each pair of border nodes, are clustered 206 into classes according
to service parameters (e.g. delay and bandwidth ranges) provided as input. The set
of specific services parameter defines a Class of Services (CoS) also known here as
CoS settings 204. For the sake of simplicity, but without any loss of generality,
in the following we consider a partition of the demands from a given border node u
to a given border node v into 6 classes associated with 6 baskets as follows:
- 3 Delay classes: DL-H (high), DL-M (medium) or DL-L (low). More in detail, each delay
class will correspond to an interval of delays, where DL stands for delay;
- 2 Bandwidth classes: BW-H (high), specified by the interval [1, b(mid)] or BW-L (low),
specified by the interval [b(mid) +1, b(max)], where BW stands for bandwidth.
[0031] The six CoS baskets are formed by all possible combinations of the three delay classes
and two bandwidth classes.
[0032] An example of such basket classification is shown in Fig. 3.
[0033] In order to build EM 208 the following assumptions are made:
- for each ordered pair (u,v) of border nodes the number of traffic demands in EM from
u to v is a same number 6K;
- in its turn, the 6K traffic demands from u to v is made of 6 sets of traffic demands
of size K, i.e. one set for each basket;
- the bandwidth of the K traffic demands in one (out of six) basket B from u to v follows
a Poisson distribution with expected value bw(u,v,B), where bw(u,v,B) is consistent
with the interval of bandwidth values for basket B. That should improve the stability
of the overall system, because there would be a sufficient number of paths of bandwidth
bw(u,v,B), which would be the more frequent value (and therefore the one that would
probably be advertised as Guaranteed, as explained in the following).
[0034] Given the expected traffic matrix EM, the solution tries to route as many demands
as possible, while minimizing the congestion on the network. Again, the goal of this
function is to improve the stability of the overall system by computing paths that,
once selected to serve a service request, would not affect other ones' resources,
minimizing the need of updating the baskets' parameters to be advertised.
[0035] This step should be performed by the domain Internal PCE (I-PCE) 210 using information
about domain topology 220 and information available from a Traffic Engineering Database,
TE DB. 222. The domain Internal PCE (I-PCE) 210 performs a successive shortest paths
algorithm, working as follows:
- the set of demands are first ordered according to some heuristic criterion that maximize
the number of successful path computations
- at each stage i, the i-th demand is routed. This will be done on a residual network
G(i), as the demands that have been routed through stages 1...i-1 have already consumed
some of the network resources. In particular, each link of G(i) will have a dynamic
congestion cost taking into account the congestion of that link so far. So, if the
i-th demand d is from u to v, the algorithm will look for a shortest path from u to
v in the network G(i) with respect to the congestion costs. Such a path will be found
via a constrained Dijkstra algorithm (the constraints guarantee that the routing paths
are consistent with the delay class of d).
[0036] For each traffic demand d in the expected traffic matrix EM that has been routed,
there is a path P that will be put in a suitable basket 212. Each path P in the basket
will have its own bandwidth (corresponding to the bandwidth of the demand of EM routed
through P by the routing algorithm) and class of delay (again corresponding to an
interval of delays). The baskets are therefore filled with the paths computed by I-PCE
according to the EM matrix, building a baskets (i.e. CoS) database 214.
[0037] Each basket (i.e. CoS) is then associated to a Summary Link (SL) connecting the Border
Nodes u and v in the Summarized Topology. Such SLs (i.e. baskets) are characterized
by the following set of service-oriented advertised 216 parameters:
- Delay class DLclass(B): this parameter is the delay class of the basket B. This parameter
is of course static;
- Bandwidth class BWclass(B): this parameter is the bandwidth class of the basket B.
This parameter is of course static;
- Guaranteed bandwidth guar(B): this parameter is chosen as the most frequent bandwidth
in the basket B. (Note that, at the beginning, this value very likely will be equal
to the expected value bw(u,v,B) of the demands in the basket B). This parameter is
dynamic;
- Peak bandwidth peak(B): this is the maximum bandwidth of the paths of the basket B
(therefore it can be different from either b(mid) or b(max) if there is no path in
the basket with such bandwidth). This parameter is a quasi-static one, because it
will be refreshed only if there is some update for the dynamic value (guar(b)). Therefore
it is possible that no path currently available in the basket has that bandwidth;
- Guaranteed bandwidth cost G-cost(B): this parameter is indicative of the (economic)
cost of selecting a path with bandwidth equal to guar(B) among the ones in the basket
B. This cost can be an economic-driven one, set by the domain administrator. For the
sake of simplicity, but without any loss of generality, from now on we will consider
the ratio between the guaranteed bandwidth and 100 times the average delay of paths
in the basket (the latter value is taken as static, i.e. it does not depend on the
paths currently available in the basket). This parameter is a dynamic one.
- Peak bandwidth cost P-cost(B): this parameter is indicative of the (economic) cost
of selecting a path with bandwidth equal to peak(B) among the ones in the basket B.
This cost can be an economic-driven one, set by the domain administrator. For the
sake of simplicity, but without any loss of generality, from now on we will consider
the ratio between the peak bandwidth and 100 times the average delay of paths in the
basket (the latter value is taken as static, i.e. it does not depend on the paths
currently available in the basket). This parameter is a quasi-static one.
[0038] To clarify the meaning of the above described parameters, let's consider an example,
where a basket B is supposed to be filled with 20 paths. In the example, the basket
B, corresponding to CoS 1, is filled with 20 paths (P1-P20) having a delay belonging
to the interval [0,5) and a bandwidth belonging to the interval [5,10], so such intervals
are advertised as DLclass(B) and BWclass(B) respectively.
[0039] Let's suppose that the bandwidth of the demands served by the considered basket B
follows a Poisson distribution with expected value equal to 8, so that 15 paths (P2-P16)
out of 20 have a bandwidth equal to 8. Let's suppose also that the path with higher
bandwidth (P1) has a bandwidth equal to 9 and a delay equal to 0.2, while the mean
delay of the paths P2-P16 is equal to 0.3. In such situation the advertised parameters
for the basket B would be the ones shown in Fig. 4.
[0040] As a final step, the requested Summarized Topology is built 218. Such summarized
topology synthesizes the domain by a full mesh whose nodes are the border nodes and
such that for each ordered pair of border nodes (u, v) there are 6 links from u to
v, each one corresponding to one basket.
[0041] Fig. 6 illustrates the method of building a summarised topology of a communications
network according to one embodiment of the present invention. As explained earlier
the method includes a step of receiving input information, 602, and this includes
CoS settings, traffic information and traffic demands for traffic between pairs of
border nodes. The input information may be received in various ways: as a combined
input with all the information is one file or package, or separately. The way the
input information is received or order in which data is input is not critical for
operation of the method. In the following step an Expected Traffic Matrix EM is built
and traffic demands represented in the Expected Traffic Matrix EM are routed, 604.
For traffic demands in the EM matrix the method comprises creating 606 a path and
assigning, 608, the path to one of CoS baskets, where the CoS baskets represent different
Classes of Service. In a preferred embodiment in steps 606 and 608 a path is created
and assigned to a CoS basket for each traffic demand in the EM matrix.
[0042] A following step comprises associating, 610, each CoS basket with a Summary Link
connecting a pair of border nodes, wherein the Summary link represents a set of connections.
Alternatively, only some of the CoS baskets are associated with their corresponding
Summary Links, e.g. when there are no connections between pair of border nodes that
would meet the CoS criteria of the CoS basket and in consequence there is no Summary
Link that could be associated with this CoS basket. This may happen, for example,
when an optical fibre connecting a pair of border nodes is broken.
[0043] Finally the method comprises building, 612, a summarised topology of the network,
wherein pairs of border nodes are linked by the Summary Links corresponding to the
CoS baskets.
[0044] Fig. 7 illustrates a device, 700, for building a summarised topology of a communications
network in accordance with one embodiment of the present invention. In one embodiment
the device is an Internal Path Computation Element (I-PCE). The device, 700, comprises
an interface, 702, adapted to receive Class of Service, CoS, settings, traffic information
and traffic demands for traffic between pairs of border nodes. The device further
comprises a processor, 704, adapted to build an Expected Traffic Matrix, EM, and route
traffic demands represented in the Expected Traffic Matrix EM. The processor, 704,
is further adapted to create a path and assign the path to one of CoS baskets for
traffic demands in the EM matrix, where the CoS baskets represent different Classes
of Service. The processor, 704, is also adapted to associate each CoS basket with
a Summary Link connecting a pair of border nodes, wherein the Summary link represents
a set of connections. The device further comprises a topology module, 706, adapted
to build a summarised topology of the network, wherein pairs of border nodes are linked
by the Summary Links corresponding to the CoS baskets.
[0045] In one embodiment the processor 704 and the topology module 706 are implemented as
separate elements and in an alternative embodiment the processor 704 and the topology
module 706 are implemented as one hardware element 708. In a preferred embodiment
the hardware element 708 is a processor having adequate processing power to realise
the functions of both elements 704 and 706.
Service & Update phase
[0046] As an example, an application of the proposed summarization method could be a PCE
hierarchy in a multi-domain context, where the parent PCE is responsible for the inter-domain
path computation on a topology, which is composed by a set of domains' summarized
topologies. In this example, the parent PCE, called External PCE (E-PCE), selects
an E2E path that is suitable for a given traffic demand, according to its service
class (i.e. the bandwidth and delay requirements of the demand) and having the minimum
cost.
[0047] Therefore, each traffic demand, which is specified by a source, a destination, and
bandwidth q, can be associated to a basket B that is compatible with its features.
A traffic demand with a set of service parameters is also known as a service demand.
[0048] If the value q is less or equal than the guaranteed bandwidth guar(B) of B, the demand
will certainly be routed. Else, it will be checked if there is some path in the basket
that has enough bandwidth to serve the demand: if not, the request will be discarded.
[0049] As shown in Fig. 5, if the demand 502 can be served, the path P 532 in the basket
530 with the smallest bandwidth greater than q will be chosen 504 and the residual
bandwidth will be made available again, as explained in the following.
[0050] The service demand will be routed on the selected path P. Therefore the path P becomes
the service path for that traffic demand and the relative basket 530 will be updated,
as shown in the Service & Update phase flow chart of Fig. 5.
[0051] After selecting a path P as the service path for a given traffic demand, the path
P will be removed from the sets of path available in the basket 530. That action can
lead to different kinds of updates 504 for the basket B from which the path P was
selected, as described in the following.
[0052] The surplus resources (if any) will be released 506 and made available again in the
TE DB 508. That can happen if, as an example, the path P was computed with a bandwidth
greater than the demand's one. A Label Switched Path, LSP, that will effectively be
setup, after the selection of path P as the service path, will reserve only the required
bandwidth (i.e. q) so the surplus will be considered available for future reservations.
[0053] In order to improve the stability of the system, the goal of the proposed summarization
method is to minimize the opportunity to change the dynamic parameters advertised
(i.e. the guaranteed bandwidth and the associated cost). To do so, the number of paths
in the basket B with a bandwidth equal or higher then Guar(B) will be monitored 510
in order to check if it is above a given threshold (e.g. set by the domain administrator).
[0054] If yes, there is no need to refill the basket B or to update its advertised dynamic
values 512.
[0055] If not, new paths to refill the basket will be searched 514. The refill will be made
along the same lines as above. Therefore, a new set of K (estimated) demands (i.e.
the refilling request 514 in the flow chart) that are compatible with the basket parameters
(i.e. in terms of bandwidth and delay) are built. Particularly, the demands will be
built with the goal of looking for paths with bandwidth equal to Guar(b) (i.e. the
ones that, if found, will produce any change in the advertised basket parameters).
[0056] Then, the K demands are routed via the I-PCE routing algorithm 516, with the constraint
that resources assigned to paths that are currently in some basket are unavailable
(note that the surplus resources, instead, are again available).
[0057] If some path with the desired characteristics is found 518, then the basket B is
refilled 520 with them and no update for its advertised dynamic values is needed,
528.
[0058] If no path with the desired characteristics is found, the basket B cannot be refilled.
Therefore, it will be checked 522 if there are still paths with a bandwidth greater
or equal to Guar(B) in the basket B.
[0059] If yes, then there is not yet the need to update the advertised dynamic values of
basket B, 512.
[0060] If no, then a bandwidth equal to Guar(B) can no more be guaranteed, and therefore
both the Guar(B) and G-cost(B) dynamic values of basket B must be updated, 524, modifying
the resources of the summarized topology of the considered domain, 526.
[0061] That way, the advertised value of Guar(B) doesn't need to be updated until the I-PCE
is no more able to find paths to refill the basket AND there are no more paths with
a bandwidth greater or equal to Guar(B) in the basket B.
[0062] It's worth to notice that the stability of the overall system is achieved through
a set of expedients performed in almost each step of the proposed summarization method.
[0063] As a first thing, the paths connecting BNs are computed according to the principle
of minimum sharing of resources (ideally no sharing at all); then the dynamic values
of each basket are chosen as the more frequent ones, taking into account that they
will be updated only in particular conditions (i.e. when some thresholds are crossed
AND there are not enough resources in the domain to guarantee the advertised characteristic
for the connections).
[0064] Fig. 8 illustrates one embodiment of a communications network, 800, for example,
an MPLS network, with an I-PCE, 700, operating and having internal structure as in
the embodiments described earlier. The network, 800, comprises a first network domain,
802, and a second network domain, 808. Each domain, 802 and 808, comprises a number
of network elements, 806, 814, for example switches and/or routers. The first domain
has a first internal path computation element (I-PCE 1), 700, connected to a first
traffic engineering database (TE DB 1), 804. The second domain 808, has a similar
structure with I-PCE 2, 810 and TE DB 2, 812.
[0065] The proposed invention provides the summarization of a domain's topology according
to a set of technology-independent service-oriented and economic-driven Class of Services
(CoS) by defining an innovative domain topology summarization method that:
- 1. provides scalability to the summarized topologies by representing a large set of
connections between a given couple of BNs as a small number of Summary Links, with
different characteristics. This way, the topology of a domain is kept small and, at
the same time, its connectivity capabilities are clustered into different Class of
Services (CoS);
- 2. provides stable summarized topologies (i.e. robust to resources status changes)
thanks to a set of sub-sequential expedients that minimize the resource sharing among
paths and re-compute specific paths with the aim of minimizing the updates to the
dynamic advertised values;
- 3. keeps confidentiality of the domain information by hiding the number, type and
TE characteristics of the intra-domain paths. As a matter of fact, only the availability
of an undefined set of connections, with some delay and bandwidth characteristics
is advertised, with no details about their composition, TE parameters, etc;
- 4. provides connections with efficient resources' utilization by providing an adequate
level of detail of the service-oriented characteristics (i.e. bandwidth and delay),
which allow the knowledge of some surely met requirements (i.e. the guaranteed bandwidth
and the E2E delay range advertised for each SL).
1. A method of building a summarised topology of a communications network comprising
the steps of:
- receiving (602) Class of Service, CoS, settings defining a plurality of CoS baskets,
each CoS basket being defined by a specific combination of a range of delay and a
range of bandwidth;
- receiving (602) traffic information;
- receiving (602) traffic demands for traffic between pairs of border nodes;
- building an Expected Traffic Matrix EM based on the received Class of Service settings,
traffic information and traffic demands, such that each traffic demand, which is specified
by a source, a destination and bandwidth, is associated to a basket compatible with
the features of said traffic demand; and routing (604) traffic demands represented
in the Expected Traffic Matrix EM;
- for traffic demands in the EM matrix creating (606) a path and assigning (608) the
path to one of CoS baskets, where the CoS baskets represent different Classes of Service,
wherein if a traffic demand can be served then a path in the basket with the smallest
bandwidth greater than the bandwidth of the traffic demand is chosen for assigning
to the traffic demand and the residual bandwidth is made available for future reservations;
- associating (610) each CoS basket with a Summary Link connecting a pair of border
nodes, wherein the Summary link represents a set of connections; and
- building (612) a summarised topology of the network, wherein pairs of border nodes
are linked by the Summary Links corresponding to the CoS baskets.
2. The method according to claim 1, wherein the traffic information is received in form
of a traffic matrix.
3. The method according to claim 1, wherein the traffic information is received in form
of stochastic distributions.
4. The method according to any one of claims 1 - 3, wherein a traffic demand with specified
a bandwidth, source and destination is assigned to a basket compatible with the features
of the traffic demand for providing the requested service by building an end-to-end
path.
5. The method according to claim 4, wherein in the assigned CoS basket a path with a
smallest bandwidth equal or greater than the specified bandwidth of the traffic demand
is selected to serve the traffic demand.
6. The method according to claim 5, wherein the remainder bandwidth of the path selected
to serve the traffic demand is available for serving another traffic demand by recording
its availability in traffic engineering database, TE DB.
7. A device (700) for building a summarised topology of a communications network comprising
an interface (702) adapted to receive Class of Service, CoS, settings defining a plurality
of CoS baskets, each CoS basket being defined by a specific combination of a range
of delay and a range of bandwidth, traffic information and traffic demands for traffic
between pairs of border nodes; a processor (704) adapted to build an Expected Traffic
Matrix EM based on the received Class of Service settings, traffic information and
traffic demands, such that each traffic demand, which is specified by a source, a
destination and bandwidth, is associated to a basket compatible with the features
of said traffic demand and route traffic demands represented in the Expected Traffic
Matrix EM, wherein the processor (704) is further adapted to create a path and assign
the path to one of CoS baskets for traffic demands in the EM matrix, where the CoS
baskets represent different Classes of Service, wherein if a traffic demand can be
served then the processor (704) is further adapted to choose a path in the basket
with the smallest bandwidth greater than the bandwidth of the traffic demand for assigning
to the traffic demand and to make the residual bandwidth available for future reservations,
the processor (704) is also adapted to associate each CoS basket with a Summary Link
connecting a pair of border nodes, wherein the Summary link represents a set of connections;
and the device further comprises a topology module (706) adapted to build a summarised
topology of the network, wherein pairs of border nodes are linked by the Summary Links
corresponding to the CoS baskets.
8. The device (700) according to claim 7, wherein the processor (704) and the topology
module (706) are implemented in the same hardware element (708).
1. Verfahren zur Bildung einer Summentopologie eines Kommunikationsnetzes, umfassend
die folgenden Schritte:
- Empfangen (602) von Dienstklassen, CoS,-Einstellungen, die eine Mehrzahl von CoS-Körben
definieren, wobei jeder CoS-Korb durch eine spezifische Kombination eines Verzögerungsbereichs
und eines Bandbreitenbereichs definiert wird;
- Empfangen (602) von Verkehrsinformationen;
- Empfangen (602) Verkehrsanforderungen für Verkehr zwischen Paaren von Randknoten;
- Bilden einer Matrix erwarteten Verkehrs EM basierend auf den empfangenen Dienstklasseneinstellungen,
Verkehrsinformationen und Verkehrsanforderungen, derart dass jede Verkehrsanforderung,
die durch einen Ursprung, ein Ziel und eine Bandbreite spezifiziert ist, mit einem
Korb assoziiert wird, der mit den Merkmalen der Verkehrsanforderung kompatibel ist;
und Weiterleiten (604) von Verkehrsanforderungen, die in der Matrix erwarteten Verkehrs
EM dargestellt sind;
- Erzeugen (606) für Verkehrsanforderungen in der EM-Matrix eines Pfades und Zuordnen
(608) des Pfades zu einem der CoS-Körbe, wobei die CoS-Körbe verschiedene Dienstklassen
darstellen, wobei dann, wenn eine Verkehrsanforderung bedient werden kann, ein Pfad
im Korb mit der kleinsten Bandbreite, die größer als die Bandbreite der Verkehrsanforderung
ist, zum Zuordnen zur Verkehrsanforderung gewählt wird, und die restliche Bandbreite
für künftige Reservierungen verfügbar gemacht wird;
- Assoziieren (610) jedes CoS-Korbes mit einer Summenverbindung, die ein Paar von
Randknoten verbindet, wobei die Summenverbindung einen Satz von Verbindungen darstellt;
und
- Bilden (612) einer Summentopologie des Netzes, wobei Paare von Randknoten durch
die Summenverbindungen, die den CoS-Körben entsprechen, verbunden werden.
2. Verfahren nach Anspruch 1, wobei die Verkehrsinformationen in Form einer Verkehrsmatrix
empfangen werden.
3. Verfahren nach Anspruch 1, wobei die Verkehrsinformationen in Form von stochastischen
Verteilungen empfangen werden.
4. Verfahren nach einem der Ansprüche 1 - 3, wobei eine Verkehrsanforderung, wobei eine
Bandbreite, ein Ursprung und ein Ziel spezifiziert sind, zu einem Korb zugeordnet
wird, der mit den Merkmalen der Verkehrsanforderung kompatibel ist, um den angeforderten
Dienst durch Bilden eines Ende-zu-Ende-Pfades bereitzustellen.
5. Verfahren nach Anspruch 4, wobei im zugeordneten CoS-Korb ein Pfad mit einer kleinsten
Bandbreite, die gleich wie oder größer als die spezifizierte Bandbreite der Verkehrsanforderung
ist, ausgewählt wird, um die Verkehrsanforderung zu bedienen.
6. Verfahren nach Anspruch 5, wobei die restliche Bandbreite des Pfades, der ausgewählt
wird, um die Verkehrsanforderung zu bedienen, durch Aufzeichnen ihrer Verfügbarkeit
in einer verkehrstechnischen Datenbank, TE-DB, zum Bedienen einer anderen Verkehrsanforderung
verfügbar ist.
7. Vorrichtung (700) zum Bilden einer Summentopologie eines Kommunikationsnetzes, umfassend
eine Schnittstelle (702), die so ausgelegt ist, dass sie Dienstklassen, CoS,-Einstellungen,
die eine Mehrzahl von CoS-Körben definieren, wobei jeder CoS-Korb durch eine spezifische
Kombination eines Verzögerungsbereichs und eines Bandbreitenbereichs definiert wird,
Verkehrsinformationen und Verkehrsanforderungen für Verkehr zwischen Paaren von Randknoten
empfängt; einen Prozessor (704), der so ausgelegt ist, dass er eine Matrix erwarteten
Verkehrs EM basierend auf den empfangenen Dienstklasseneinstellungen, Verkehrsinformationen
und Verkehrsanforderungen bildet, derart dass jede Verkehrsanforderung, die durch
einen Ursprung, ein Ziel und eine Bandbreite spezifiziert ist, mit einem Korb assoziiert
wird, der mit den Merkmalen der Verkehrsanforderung kompatibel ist; und Verkehrsanforderungen
weiterleitet, die in der Matrix erwarteten Verkehrs EM dargestellt sind, wobei der
Prozessor (704) ferner so ausgelegt ist, dass er einen Pfad bildet und den Pfad zu
einem der CoS-Körbe für Verkehrsanforderungen in der EM-Matrix erzeugt, wobei die
CoS-Körbe verschiedene Dienstklassen darstellen, wobei der Prozessor (704) ferner
so ausgelegt ist, dass er dann, wenn eine Verkehrsanforderung bedient werden kann,
einen Pfad im Korb mit der kleinsten Bandbreite, die größer als die Bandbreite der
Verkehrsanforderung ist, zum Zuordnen zur Verkehrsanforderung wählt und die restliche
Bandbreite für künftige Reservierungen verfügbar macht, wobei der Prozessor (704)
außerdem so ausgelegt ist, dass er jeden CoS-Korb mit einer Summenverbindung assoziiert,
die ein Paar von Randknoten verbindet, wobei die Summenverbindung einen Satz von Verbindungen
darstellt; und die Vorrichtung ferner ein Topologie-Modul (706) umfasst, das so ausgelegt
ist, dass es eine Summentopologie des Netzes bildet, wobei Paare von Randknoten durch
die Summenverbindungen, die den CoS-Körben entsprechen, verbunden werden.
8. Vorrichtung (700) nach Anspruch 7, wobei der Prozessor (704) und das Topologie-Modul
(706) im gleichen Hardwareelement (708) implementiert sind.
1. Procédé de construction d'une topologie synthétique d'un réseau de communication,
comprenant les étapes de :
- la réception (602) de réglages de classe de service, CoS, définissant une pluralité
de paniers de CoS, chaque panier de CoS étant défini par une combinaison spécifique
d'une plage de retard et d'une plage de largeur de bande ;
- la réception (602) d'informations de trafic ;
- la réception (602) de demandes de trafic pour un trafic entre des paires de noeuds
de frontière ;
- la construction d'une matrice de trafic prévu, EM, sur la base des réglages de classe
de service reçus, des informations de trafic reçues et des demandes de trafic reçues,
de sorte que chaque demande de trafic, qui est spécifiée par une source, une destination
et une largeur de bande, soit associée à un panier compatible avec les caractéristiques
de ladite demande de trafic ; et l'acheminement (604) de demandes de trafic représentées
dans la matrice de trafic prévu, EM ;
- pour des demandes de trafic dans la matrice EM, la création (606) d'un trajet et
l'assignation (608) du trajet à l'un des paniers de CoS, dans lequel les paniers de
CoS représentent différentes classes de service, dans lequel, si une demande de trafic
peut être desservie, alors un trajet dans le panier avec la plus petite largeur de
bande supérieure à la largeur de bande de la demande de trafic est choisi pour son
assignation à la demande de trafic et la largeur de bande résiduelle est mise à la
disposition de futures réservations ;
- l'association (610) de chaque panier de CoS à une liaison synthétique reliant une
paire de noeuds de frontière, dans lequel la liaison synthétique représente un ensemble
de connexions ; et
- la construction (612) d'une topologie synthétique du réseau, dans lequel des paires
de noeuds de frontière sont reliées par les liaisons synthétiques correspondant aux
paniers de CoS.
2. Procédé selon la revendication 1, dans lequel les informations de trafic sont reçues
sous forme d'une matrice de trafic.
3. Procédé selon la revendication 1, dans lequel les informations de trafic sont reçues
sous forme de distributions stochastiques.
4. Procédé selon l'une quelconque des revendications 1 - 3, dans lequel une demande de
trafic avec une largeur de bande spécifiée, une source spécifiée et une destination
spécifiée est assignée à un panier compatible avec les caractéristiques de la demande
de trafic pour fournir le service demandé par la construction d'un trajet de bout
en bout.
5. Procédé selon la revendication 4, dans lequel, dans le panier de CoS assigné, un trajet
avec la plus petite largeur de bande supérieure ou égale à la largeur de bande spécifiée
de la demande de trafic est sélectionné pour desservir la demande de trafic.
6. Procédé selon la revendication 5, dans lequel la largeur de bande restante du trajet
sélectionné pour desservir la demande de trafic est disponible pour desservir une
autre demande de trafic en enregistrant sa disponibilité dans une base de données
d'ingénierie de trafic, TE DB.
7. Dispositif (700) de construction d'une topologie synthétique d'un réseau de communication,
comprenant une interface (702) apte à effectuer la réception de réglages de classe
de service, CoS, définissant une pluralité de paniers de CoS, chaque panier de CoS
étant défini par une combinaison spécifique d'une plage de retard et d'une plage de
largeur de bande, et la réception d'informations de trafic et de demandes de trafic
pour un trafic entre des paires de noeuds de frontière ; un processeur (704) apte
à effectuer la construction d'une matrice de trafic prévu, EM, sur la base des réglages
de classe de service reçus, des informations de trafic reçues et des demandes de trafic
reçues, de sorte que chaque demande de trafic, qui est spécifiée par une source, une
destination et une largeur de bande, soit associée à un panier compatible avec les
caractéristiques de ladite demande de trafic, et l'acheminement de demandes de trafic
représentées dans la matrice de trafic prévu, EM ; dans lequel le processeur (704)
est en outre apte à effectuer la création d'un trajet et l'assignation du trajet à
l'un des paniers de CoS pour des demandes de trafic dans la matrice EM, dans lequel
les paniers de CoS représentent différentes classes de service, dans lequel, si une
demande de trafic peut être desservie, alors le processeur (704) est en outre apte
à choisir un trajet dans le panier avec la plus petite largeur de bande supérieure
à la largeur de bande de la demande de trafic pour son assignation à la demande de
trafic et mettre la largeur de bande résiduelle à la disposition de futures réservations
; le processeur (704) est également apte à effectuer l'association de chaque panier
de CoS à une liaison synthétique reliant une paire de noeuds de frontière, dans lequel
la liaison synthétique représente un ensemble de connexions ; et le dispositif comprend
en outre un module de topologie (706) apte à effectuer la construction d'une topologie
synthétique du réseau, dans lequel des paires de noeuds de frontière sont reliées
par les liaisons synthétiques correspondant aux paniers de CoS.
8. Dispositif (700) selon la revendication 7, dans lequel le processeur (704) et le module
de topologie (706) sont mis en oeuvre dans le même élément matériel (708).